Proceedings Fonetik 2009 - Institutionen för lingvistik

Proceedings, FONETIK 2009, Dept. of Linguistics, Stockholm University• Female sounding male | male looking female| undecided (when voice♂ & face♀)• Male sounding female | female lookingmale | undecided (when voice♀ & face♂)• Natural | dubbed | undecided (when speakercongruent, vowel incongruent)Upon completing the responses, these weretransmitted by e-mail to the experimenter, togetherwith possible comments by the subject,who was invited to an explanatory demonstration(http://legolas.ling.su.se/staff/hartmut/webexperiment/xmpl.se.htm, ...tk.htm).Subjects participating via Swedish web forawere informed within 15 minutes or so abouthow many times they had heard by eye.ResultsThe most essential stimulus specific results aresummarized in Table 1 for Exp. 1, 2 and 4 andin Table 2 for Exp. 3. Subjects who had not indicatedthe relevant language as their first ortheir best known language have been excluded.It can be seen in Table 1 and 2 that for eachauditory-visual stimulus combination, therewere only minor differences in the results betweenExp. 1, 2, and the Swedish version ofExp. 4. For combinations of auditory [e] andvisual [y], the influence of vision was, however,clearly smaller than that observed within theframe of the previous experiment (Traunmüllerand Öhrström, 2007), while it was clearlygreater in the Turkish version of Exp. 4. Absenceof lip rounding in the visible face hadgenerally a stronger effect than visible presenceof lip rounding, in particular among Swedes. Inthe Turkish version, there was a greater influenceof vision also for the combination of auditory[y] and visual [e], which was predominantlyperceived as an [i] and only seldom as [yj]among both Turks and Swedes. The response[ɯ] (Turkish only) was also rare. The proportionof visually influenced responses substantiallyexceeded the proportion of stimuli perceivedas undubbed, especially so amongTurks.The results from Exp. 3, in which thespeakers had been switched (Table 2), showedthe same trend that can be seen in Exp. 1, althoughvisible presence of roundedness had aprominent effect with the female speaker withinthe frame of the previous experiment.A subject-specific measure of the overall influenceof vision was obtained by counting theresponses in which there was any influence ofvision and dividing by four (the number of incongruentstimuli presented).A preliminary analysis of the results fromExp. 1 to 3 did not reveal any substantial effectsof habituation to dubbing, discomfort fromdubbing, sex or age.Table 2.Summary of stimulus specific results forExp. 3 arranged as in Table 1.StimulusVoice & FacePrev.exp.n=42,20 ♂, 22 ♀ OExp 3n=4741 ♂, 6 ♀Naty ♂ & e ♂ 79 4 81 66e ♀ & y ♀ 86 3 34 11y ♀ & e ♂ - 2 85e ♂ & y ♀ - 1 28Table 1. Summary of stimulus specific results from Exp. 1, 2, and 4: Percentage of cases showing influenceof vision on heard roundedness in syllable nucleus (monophthong or diphthong). No influence assumedwhen response was ‘undecided’. O: Order of presentation. Nat: Percentage of stimuli perceived as natural(undubbed) shown for stimuli without incongruence in sex. Corresponding results from previous experiment(Traunmüller and Öhrström, 2007) shown in leftmost column of figures.StimulusVoice & FacePrev.exp.n=42,20 ♂,22 ♀ OExp. 1n=185122 ♂,63 ♀ Nat OExp. 2Informedn=99,57 ♂,42 ♀ Nat OExp. 4Swedsn=84,73 ♂,11 ♀ NatExp. 4Turks,n=71,30 ♂,41 ♀ Naty ♀ & e ♀ 83 4 82 72 4 81 66 3 81 61 99 64e ♂ & y ♂ 50 3 26 15 3 25 11 4 23 9 79 23y ♂ & e ♀ - 2 80 2 65 1 75 94e ♀ & y ♂ - 1 41 1 42 2 52 83169